Mapping Agricultural Tillage Practices Using Extreme Learning Machine

Abstract

In this paper, an efficient classifier based on extreme learning machine (ELM) is proposed to use for mapping agricultural tillage practices from hyperspectral remote sensing imagery. The kernel version, called kernel ELM (KELM), is implemented due to its powerfulness. To utilize spatial information of an image, a spatial convolution filter is adopted to generate spatial-spectral features of a hyperspectral pixel by incorporating its surrounding pixels, which are the actual inputs to the KELM. Experimental results using airborne hyperspectral images demonstrate that the KELM can outperform other classic methods, such as support vector machine and random forest.